Conversational Papers (cPAPERS)
- Type
- dataset
- Venue
- Hugging Face / AVA Lab
- Year
- 2026
- Source
- huggingface
- Access
- free
- Language
- English
- Added
- 2026-07-17T20:18:03.678532+00:00
- Verified
- 2026-07-17T20:18:03.678532+00:00
Summary
cPAPERS is a dataset of conversational question-answer pairs grounded in scientific paper components, published at NeurIPS 2024. It contains QA pairs pertaining to figures (cPAPERS-FIGS), equations (cPAPERS-EQNS), and tabular information (cPAPERS-TBLS) extracted from academic papers. Question-answer pairs are sourced from OpenReview reviews and rebuttals and associated with contextual information from arXiv LaTeX source files, including surrounding text, references, and LaTeX-formatted equations. The dataset is designed to support the development of conversational assistants capable of interactive discussions about scientific papers.
Keywords
scientific-papers qa conversational-ai simmc figures equations tables openreview arxiv neurips
Topics
NLP / Scientific Papers
Research notes
- Published by AVA Lab (Georgia Tech) on HuggingFace. Paper: arXiv:2406.08398, NeurIPS 2024. GitHub: avalab-gt/cPAPERS. QA pairs extracted using LLaMA+GPT pipeline. The HF dataset viewer has a cast error due to mismatched columns between cFIGS and cEQNS subsets.